2015 Great Lakes Region

Started by lastguyoffthebench, September 07, 2015, 12:56:39 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Domino1195

I think, as with previous years, SOS is playing a large role in the rankings. NCAC earlier lamented the weakness of Kenyon's schedule - 28th of the 52 midwest teams, lowest of the 8 ranked teams.


It will be interesting to watch how SOS changes and what teams do to other's rankings by virtue of how well the finish - or falter - in the next two weeks.  Capital, for example, can impact Denison, OWU, TM, Case and ONU:



InstitutionIn-Division SOS (weighted OWP-OOWP)
Wittenberg
0.583[/t][/t] 
Allegheny
0.58[/t] 
Denison
0.578[/t] 
Ohio Wesleyan
0.564[/t] 
Oberlin
0.551[/t] 
Rose-Hulman
0.549[/t] 
Wilmington (OH)
0.547[/t] 
John Carroll
0.546[/t] 
DePauw
0.544[/t] 
Carnegie Mellon
0.541[/t] 
Geneva
0.539[/t] 
Anderson (IN)
0.537[/t] 
Grove City
0.535[/t] 
Hiram
0.535[/t] 
Capital
0.533[/t] 
Heidelberg
0.528[/t] 
CWRU
0.523[/t] 
Marietta
0.52[/t] 
Otterbein
0.517[/t] 
Thomas More
0.516[/t] 
Wabash
0.511[/t] 
Waynesburg
0.51[/t] 
Westminster (PA)
0.508[/t] 
Ohio Northern
0.506[/t] 
Manchester
0.501[/t] 
Franklin
0.497[/t] 
Wash. & Jeff.
0.495[/t] 
Kenyon
0.494[/t] 
Earlham
0.493[/t] 
Wooster
0.492[/t] 
Hanover
0.485[/t] 
Mt. St. Joseph
0.485[/t] 
Pitt.-Greensburg
0.48[/t] 
Bethany (WV)
0.478[/t] 
Medaille
0.474[/t] 
Transylvania
0.474[/t] 
Franciscan
0.471[/t] 
D'Youville
0.463[/t] 
Mount Union
0.462[/t] 
Pitt.-Bradford
0.458[/t] 
Baldwin Wallace
0.454[/t] 
Saint Vincent
0.454[/t] 
Berea
0.452[/t] 
Penn St.-Behrend
0.447[/t] 
La Roche
0.446[/t] 
Muskingum
0.445[/t] 
Defiance
0.435[/t] 
Thiel
0.43[/t] 
Hilbert
0.425[/t] 
Mount Aloysius
0.414[/t] 
Bluffton
0.397[/t] 
Penn St.-Altoona
0.391[/t]

PaulNewman

Doesn't compute for me.  Kenyon has PLAYED 9 of the teams with SOS HIGHER than them, INCLUDING Waynesburg as one of those HIGHER!

Flying Weasel

The data is only to include games through Sunday as I understand it.  The dates on the data sheets can be and historically have been all over the place between Monday and Wednesday, varying by region and gender, which has always given me the impression it's a date/time stamp for when it was last edited, not the cut-off date for inclusion of results.  As far as I know the data sheets have not historically included results through to the date/time stamped on them. 

Do you think you are seeing records that are only correct if including results from Tuesaday?

TennesseeJed

Based on all the info available from this site, d3 soccer and NCAA, I'm still scratching my head.  If you take the SoS's of each team and multiply it by their respective in-division WL%, you should get a reasonably close proxy to the actual rankings, given that the RvR criteria hasn't yet been applied, unless I'm really missing the boat entirely.  Or, the rankings must be subject to highly weighted, qualitative factors which are not disclosed.

Here's a quick calc on the SoS * W/L% for teams in the region that have been discussed or ranked as strong regional contenders in the GL region, ranked in order of their (SoS) x (W/L%) multiple:

Thomas More      0.516   -0.893       (0.4608)
Kenyon         0.494   -0.917       (0.4530)
Ohio Wesleyan           0.564   -0.8           (0.4512)
Denison              0.578   -0.769       (0.4445)
DePauw         0.544   -0.808       (0.4396)
CWRU         0.523   -0.821       (0.4294)
Rose-Hulman      0.549   -0.75               (0.4118)
Ohio Northern      0.506   -0.813            (0.4114)
Penn St.-Behrend   0.447   -0.9               (0.4023)
John Carroll      0.546   -0.731       (0.3991)
Carnegie Mellon           0.541   -0.731       (0.3955)


If in-division w/l% and SoS are both equally weighted (which they may not be), then the multiplication of the two factors will give the best indication of record and strength of schedule across competitors within the region, ignoring other factors, such as RvR, etc., which are not yet in these numbers.  The ranking above squares much better with my sense of regional strength, particularly when the availability of SoS is transparent and provided using comparable methodology for all teams.  The actual rankings put out by NCAA today continue to baffle me, unless the SoS has a significantly higher weighting than W/L%, in which case, teams would be ranked strictly using SoS, as follows, and which does also not line up with the published rankings...

Denison              0.578
Ohio Wesleyan           0.564
Rose-Hulman           0.549
John Carroll           0.546
DePauw              0.544
Carnegie Mellon           0.541
CWRU              0.523
Thomas More           0.516
Ohio Northern           0.506
Kenyon              0.494
Penn St.-Behrend   0.447

Obviously, neither approach gets close to the actual rankings.  I'm obviously missing something, but what?

TennesseeJed

Quote from: Flying Weasel on October 21, 2015, 03:54:23 PM
The data is only to include games through Sunday as I understand it.  The dates on the data sheets can be and historically have been all over the place between Monday and Wednesday, varying by region and gender, which has always given me the impression it's a date/time stamp for when it was last edited, not the cut-off date for inclusion of results.  As far as I know the data sheets have not historically included results through to the date/time stamped on them. 

Do you think you are seeing records that are only correct if including results from Tuesaday?

From the top of the GL Regional Sheet.  Source:  NCAA Regional Rankings links at bottom of rankings page.

GREAT LAKES REGION
D-III Men's Soccer
Generated 10/19/2015 01:18 PM



The GL .pdf is stamped as of Friday, 10/19/15, yet the records must be accurate through Saturday, 10/20 at least, because for the teams I've checked, who all played Saturday games, the number of games is accurate through Saturday's games, not as of Friday or before.

chelseafc30

I just don't see how you can possibly justify Denison being ranked over Kenyon. I understand that SOS is a big deal and Denison has a much better stat (.58 vs .495). But if you actually go and LOOK at the two teams schedules, I don't see how Denison's SOS is so much higher than Kenyon's. They have a relatively similar schedule. The main differences are Denison has a win over Emory and Thomas More while Kenyon has a win vs Carnegie Mellon. I consider Emory and CMU pretty equal, so does that ONE win vs Thomas More really change a SOS from .495 to .58? Not to even mention that Kenyon just handled Denison 3-0 on Saturday, holding the Big Red to their season low 4 shots in the game.

Ryan Harmanis

#81
Quote from: TennesseeJed on October 21, 2015, 04:05:48 PM
The GL .pdf is stamped as of Friday, 10/19/15, yet the records must be accurate through Saturday, 10/20 at least, because for the teams I've checked, who all played Saturday games, the number of games is accurate through Saturday's games, not as of Friday or before.

Your dates are off.  Yesterday was 10/20, and those results are not included.  OWU is now 12-2-2, DePauw is 9-2-3, but the data sheet has them as 11-2-2 and 9-1-3.  So if they're accurate through 10/19 that would include the weekend's results, but not this week's results.

I don't disagree with anyone on this, but as FW pointed out this is pretty consistent with what the regional committees have done in the past.  Whitworth, which is still undefeated, is not ranked at all because of their low SOS, so Kenyon might be lucky (in this twisted sense) to be included despite an SOS below .500.  Now, this should take care of itself if Kenyon just needs to get above the .500 threshold, because a game with OWU plus 1-2 conference tourney games should bump it the .006 needed to get there.

And out of curiosity, does anyone have suggestions on improving this thing?  The weighted SOS*Win% is an idea, anything else?  For all its flaws, I think we'd all prefer something involving numbers rather than just a committee decision without any limitations or requirements to follow the data.  This isn't the college football playoff committee, these are the coaches of teams involved.

Flying Weasel

#82
Quote from: NCAC New England on October 21, 2015, 02:53:02 PM
Doesn't compute for me.  Kenyon has PLAYED 9 of the teams with SOS HIGHER than them, INCLUDING Waynesburg as one of those HIGHER!

But playing teams with high SOS doesn't make your SOS high.  Only 2/3 of an opponents SOS (the OWP portion) factors into 1/3 of their contribution to your SOS (the OOWP portion).  Waynesburg's record is 1-11-1 (.115).  Assuming Waynesburg's OWP is similar to their SOS, Waynesburg's contribution to Kenyon's SOS goes like this:  1.25 (away game multiplier) x [ 2/3(.115 OWP) + 1/3(.510 assumed OOWP)] = 0.308.  To offset such a poor opponent, you might need a couple good opponents. 

And that reminds me of another of the flaws in the NCAA's SOS calculations.  There should be a threshhold on how low of a win pct. is applied.  For an NCAA tournament aspiring team, there is no real difference between playing a 1-11-1 (.115) Waynesburg and a 3-7-1 (.318) Mt. St. Joseph, both are cupcakes for a Top 25-caliber team.  But one can hurt your SOS, the other can kill it.  I don't think that's right.  You'd have to experiement to home in one a good lower bound, but any opponent's winning percentage below .400  (or whatever) should get counted as .400 (or whatever) for SOS calculation purposes.  A cupcake is a cupcake, here shouldn't be degrees of cupcakes.

TennesseeJed

#83
Quote from: Ryan Harmanis on October 21, 2015, 04:12:15 PM
Quote from: TennesseeJed on October 21, 2015, 04:05:48 PM
The GL .pdf is stamped as of Friday, 10/19/15, yet the records must be accurate through Saturday, 10/20 at least, because for the teams I've checked, who all played Saturday games, the number of games is accurate through Saturday's games, not as of Friday or before.

Your dates are off.  Yesterday was 10/20, and those results are not included.  OWU is now 12-2-2, DePauw is 9-2-3, but the data sheet has them as 11-2-2 and 9-1-3.  So if they're accurate through 10/19 that would include the weekend's results, but not this week's results.

Sorry for the mistake and thanks for the correction RH.  So, the weekend's games were included for SoS and W/L%, as they should have been and time-stamped appropriately.  Sorry, again, for the confusion...

Mid-Atlantic Fan

Quote from: Flying Weasel on October 21, 2015, 04:21:12 PM
Quote from: NCAC New England on October 21, 2015, 02:53:02 PM
Doesn't compute for me.  Kenyon has PLAYED 9 of the teams with SOS HIGHER than them, INCLUDING Waynesburg as one of those HIGHER!

But playing teams with high SOS doesn't make your SOS high.  Only 2/3 of an opponents SOS (the OWP portion) factors into 1/3 of their contribution to your SOS (the OOWP portion).  Waynesburg's record is 1-11-1 (.115).  Assuming Waynesburg's OWP is similar to their SOS, Waynesburg's contribution to Kenyon's SOS goes like this:  1.25 (away game multiplier) x [ 2/3(.115 OWP) + 1/3(.510 assumed OOWP)] = 0.308.  To offset such a poor opponent, you might need a couple good opponents. 

And that reminds me of another of the flaws in the NCAA's SOS calculations.  There should be a threshhold on how low of a win pct. if applied.  For an NCAA tournament aspiring team, there is no real difference between playing a 1-11-1 (.115) Waynesburg and a 3-7-1 (.318) Mt. St. Joseph, both are cupcakes for a Top 25-caliber team.  But one can hurt your SOS, the other can kill it.  I don't think that's right.  You'd have to experiement to home in one a good lower bound, but any winning percentages below .400  (or whatever) get counted as .400 for SOS calculation purposes.

I like that idea you mentioned above in bold/italics. I doubt teams that are trying to make the tournament want to play a 1-15-1 team. That same team was probably an average just below .500 record the year before and for some reason tanked. It's unpredictable sometimes and that can affect a team's chances. An example of this that I can think of this year comes from the Empire 8 league with Elmira. They are 6-4-4 this year as compared to last year being 1-14-2. I doubt many people wanted to pick them up but for the teams that did they actually got rewarded in a way for it with the drastic progress that Elmira has shown in their turn around this season. At the same time for the teams that are picking other teams that they know will have a good record and be a challenge it's kind of unfair because they don't get any "bonus" points for playing a team with a good SOS. It's a good idea but tough to implement.

PaulNewman

#85
And Case is about to play Waynesburg!  I suggested they drop that and play the Kenyon game but I'm sure that is impossible as well as probably just wrong.  And there is no way for teams to know ahead of the year when they schedule.  As I noted a while back Waynesburg actually was pretty good last year.  I don't have an answer but when you line up the schedules of two teams and they look remarkably similar I just don't think we can accept such divergence.  Obviously the Kenyon fans are outraged today. If it was reversed and this was happening to a 1 loss (in OT) OWU team I'm sure they would be outraged.  And the point about Denison....my gosh, they just lost to Allegheny and HEAD-TO-HEAD vs Kenyon in a mismatch.  Let's be real here.  Kenyon was ranked #1 in both polls just 3 weeks ago.  Now they can't break the top 5 in their own region?  And the only thing that happened since on the negative end was an OT loss in a game they dominated against a ranked team?

I do agree about the Case game, and as I sort of foreshadowed the Case game likely would cost them TWO games...the Case game and the loss to DPU in a game they thoroughly dominated but were still not entirely prepared for.

Ryan Harmanis

Figured if we want to continue the Kenyon-ranking thread of conversation it might make more sense to do so here.  I'm happy to continue the discussion as I think it has merit, but I don't want to hijack the national thread. 

chelseafc30

I agree completely with RH. No need to hijack the the national thread with Kenyon, considering nearly every post is dealing with Kenyon's conundrum of a regional rank. That being said, I would like to read RH's (or anyone's) justification for Denison being ranked ahead of Kenyon.

Ryan Harmanis

#88
Perhaps not a justification, but here's possible reasoning if I'm the committee, loosely using their criteria.  (1) Denison has a good record/win%; (2) Denison has really good wins over Emory and Thomas More and a tie against region #1 OWU; (3) Denison has, by far, the highest SOS of the ranked teams, and it's light-years beyond Kenyon's right now; (4) Denison and Kenyon have identical 5-0 records against common opponents (Heidelberg, Marietta, Muskingum, Wooster, Hiram).

I get that Kenyon easily handled Denison, but we can't just go head-to-head for the regional rankings.  It's circular - Denison beat Thomas More beat OWU beat DePauw beat Kenyon beat Denison.  So once you look beyond the head-to-head, it becomes less surprising, especially given Kenyon's low SOS.  I said this in the other thread, but knowing Kenyon's SOS is below .500 makes it more surprising that they were ranked at all.  Kenyon has avoided blemishes, which does count for something, but that's partially because they've played few quality teams, and the committee isn't going to speculate as to what they would have done against a better schedule.

As for the fact that Waynesburg kills the SOS, you can avoid some of the risk by upping the average level of your opponent.  When you schedule several above-average teams, as Denison did, you can afford more blemishes because your SOS is so much higher.  What's killing Kenyon isn't Waynesburg, it's the low number of non-conference games and playing below-average teams for nearly all of them.  To put it in perspective, even if you substitute Waynesburg into Denison's and OWU's schedules for Thomas More - the biggest shift possible - both teams still have a much higher non-conference opponents' win%, with Denison at 0.525 and OWU at 0.516, compared to Kenyon's 0.452.  And that's not even including Denison's cancelled game against John Carroll, which offsets (valid) complaints about Kenyon missing out on the Case game.

I just want to re-emphasize this doesn't reflect my view of the teams, or (to your point) the actual result between them.  What it does reflect is that the committee works with data, and the criteria have been known and used for a long, long time,  so Kenyon was on notice this could happen.  Frankly, I'm guessing it's what happened in 2007, when Kenyon went 14-2-2, won the NCAC, played a very weak non-conference schedule and didn't get an at-large bid when they lost in PKs in the conference semis.  That, above all else, just reinforces that Kenyon still has total control of their destiny this year with regards to an NCAA bid.  Win the NCAC tournament and we've all just been spinning our wheels here.

TennesseeJed

#89
Quote from: chelseafc30 on October 21, 2015, 04:08:46 PM
I just don't see how you can possibly justify Denison being ranked over Kenyon. I understand that SOS is a big deal and Denison has a much better stat (.58 vs .495). But if you actually go and LOOK at the two teams schedules, I don't see how Denison's SOS is so much higher than Kenyon's. They have a relatively similar schedule. The main differences are Denison has a win over Emory and Thomas More while Kenyon has a win vs Carnegie Mellon. I consider Emory and CMU pretty equal, so does that ONE win vs Thomas More really change a SOS from .495 to .58? Not to even mention that Kenyon just handled Denison 3-0 on Saturday, holding the Big Red to their season low 4 shots in the game.

The biggest differences in the two schedules are in the home vs. away differences and how that will impact each team's SoS.  Denison played 8 of its 14 games away.  Kenyon played 8 of its 13 games at home.  It's not the case in reality, but if we assumed that the OWP's and OOWP's for each team were the same for Kenyon and Denison (which would be the case if they played the exact same schedule) for all games, there could be as much as a 47% difference in the weighted/adjusted OWP for Kenyon and Denison, depending on where each game is played.  If Kenyon played a .600 team at home, they would add an adjusted OWP of .510 for that team/game.  If Denison played the exact same team, but played it as an away game, their adjusted OWP for that same team/game would be .720.  Obviously, with SoS's being a critical [primary] factor in rankings, and with the relevant spread for most ranked teams centering in the mid-.500 range (.525 to .575 seems to be the bulk of the distribution--this is an eyeball guess, not mathematically derived), the difference in adding a .510 vs. a .720 has a massive impact for something as minor (in my judgment--the NCAA obviously disagrees) as home field advantage. 

The NCAA is not just rewarding you for playing away games, they are actually penalizing you for playing games at home.  If you want to raise your SoS, the easiest way is to play more away games during the season.  It's worth close to 47% on each OWP and OOWP, so the incentive to play a game away is huge.  If your existing SoS is based on 10 home games vs. teams who all had .500 OWP's, your SoS would be roughly .425.  If you played the same  10 games away, your SoS would be roughly .600.  If you played half at home and half away, your SoS would be roughly .5125, so you can see that the NCAA strongly rewards away games, otherwise, an equal number of home and away games would offset each other in an SoS calc, assuming the OWPs and OOWPs of the teams played were the same.

The difference between Kenyon's and Denison's SoS is 1-(.578/.494) = 17%.  Denison has played 57% of its games away; Kenyon has played only 38% of its games away.  If the two schools played 3 of the same teams, but Kenyon played them all at home and Denison played them all on the road, a 19% difference times a 47% weight difference could easily reconcile a big portion (roughly 9% of the 17%) of the difference in respective SoS's... the other 8% is likely from differences in teams played and respective OWP's and OOWP's.

The main point I'm making is that with a variation in SoS's that's only about 10% wide (.575-.525/.525=.095, or 9.5%), but a multiplier that's 40%+ wide, the weighting of the home vs. away is too large and it overwhelms the underlying OWP and OOWP calcs in a way that creates very high SoS's for teams (any team--not just Kenyon or Denison) that have more away games in their schedules and very low SoS's for teams that play a lot of home games.

Note, that none of this addresses the Win % advantage Kenyon has over Denison, which needs to be accounted for in the overall ranking.  The conclusion you must draw is that the committee strongly favors SoS over win % as the relevant ranking metric, even though it doesn't capture anything about a team's performance.  It also doesn't consider the direct head to head match up or either team's results vs. ranked teams (RvR), but there is a mechanism for incorporating that information next week, if I understand the process correctly.